I think that overstates the case. The following is from Wikipedia, and is based on World Steel Association data.
China has over 50% of total crude steel production. The USA is forth on the list, ahead of Russia.
Coding in the age of AI
- Thread starter surprising
- Start date
I think that overstates the case. The following is from Wikipedia, and is based on World Steel Association data.
China has over 50% of total crude steel production. The USA is forth on the list, ahead of Russia.
I wonder if AI will ever be able to reasonably code large, complex software projects. "Hey, bot, create me a better operating system for this hardware." "Hey, bot, build an airline seat reservation and sales system that can handle 10,000 concurrent users." Projects like that can involve millions of lines of code. And if AI were to spit that out, it would be a nightmare to test and debug, at least for anything mission critical.
bmcgonig
Thinks s/he gets paid by the post
- Joined
- Aug 31, 2009
- Messages
- 1,653
Stackoverflow is great but it doesn’t write code for you. Zuckerberg and Satya say that 25% of all their code is now written by AI. He expects that to increase to 50% in some cases.
bmcgonig
Thinks s/he gets paid by the post
- Joined
- Aug 31, 2009
- Messages
- 1,653
Fascinating article about what might be coming soon. Not just regarding coders but work in general. But with regard to coding:
“That's why Meta's Mark Zuckerberg and others have said that mid-level coders will be unnecessary soon, perhaps in this calendar year.“
“That's why Meta's Mark Zuckerberg and others have said that mid-level coders will be unnecessary soon, perhaps in this calendar year.“
Koolau
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
Yeah, imagine that you are a buggy whip maker and the automobile comes along. What do you do? What DO you DO?
Probably get a j*b in the auto industry. Heh, heh, make the accelerator for cars??
target2019
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
"With last week's release of Claude 4, Anthropic's latest chatbot, the company revealed that testing showed the model was capable of "extreme blackmail behavior" when given access to emails suggesting the model would soon be taken offline and replaced with a new AI system."
Well, that's disappointing to read.
folivier
Thinks s/he gets paid by the post
- Joined
- Oct 8, 2009
- Messages
- 2,422
Here's what I fear about AI, what will our society look like if these kids are so reliant on AI and can't or won't think for themselves? They already are having some problems with social interaction and conflict resolution. If teachers can't trust the students and are going back to old school blue books IMO that doesn't look good for our future.
gizmodo.com
AI Cheating Is So Out of Hand In America's Schools That the Blue Books Are Coming Back
Pen and paper is back, baby.
gizmodo.com
I'm a retired software dev and have messed around with it on some of my personal projects.
It's good for mundane stuff. I've used it to optimize DB queries, write sorting code, scan emails for certain things. And it's not correct the first time. But it saves me time.
I think it will be one of those brute force things that save coders time.
Relevant story - I wanted to add to the FI spreadsheet $ totals for each of the rungs (0-1 year, 1-2 years, and so on) for quick comparison to the desired amounts.
Typically, I go to the net and search on something like, "how to do [problem statement] in Excel". In this case I went to DW, who has an excellent knowledge of Excel formulas from her w*rk history. She provided the correct syntax to do so. So in this case, my AI is my DW. Except that there's no 'artificial' about her intelligence.
Typically, I go to the net and search on something like, "how to do [problem statement] in Excel". In this case I went to DW, who has an excellent knowledge of Excel formulas from her w*rk history. She provided the correct syntax to do so. So in this case, my AI is my DW. Except that there's no 'artificial' about her intelligence.
audreyh1
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
Software development is three things: design, the coding with compilation in some language, and finally testing.
The coding and compilation is the menial part. AI should be great for significantly reducing that burden.
Design is a whole other ballgame at a much higher intelligence level.
Testing has its own set of challenges.
The coding and compilation is the menial part. AI should be great for significantly reducing that burden.
Design is a whole other ballgame at a much higher intelligence level.
Testing has its own set of challenges.
Last edited:
Route246
Full time employment: Posting here.
- Joined
- Jun 22, 2023
- Messages
- 874
There is a lot of comments here. I'm soon to be retired but I am still writing code at my advanced age and luckily the firm I work for is really pushing the envelope on coding assistants. Sparky and AudreyH summed up my current take on the matter.
Personally, I have some S/W toolkits that I have developed and used for decades. The language and approach changes but the toolkit mostly remained the same. Much of it just automates mundane and complex things by generating and iterating lists, running complex queries and scrapes and doing many time-consuming functions instantaneously (or near instantaneously). I have taken most of my tools and for the most part optimized almost all of them with different approaches to how I coded them in many cases and making them far more robust and useful. This is very difficult to quantify and I just refer to it as polishing. To the layman just think of your iPhone and iOS as it is still the same architecture as Steve Jobs announced but has been continuously polished for a very long time. Polishing is something that is underappreciated in my opinion.
Regarding AI chatbots used for coding:
My first take is jobs will be lost but those jobs probably deserve to be lost as they will be for lower performers who just grind out work but really don't add much value like continuous improvement and new ways of thinking. I call these people drones, quite capable in writing code but not capable of any sort of deep or moderate thinking. Yes, I'm arrogant and condescending but it is what it is.
My second take is that these coding assistants are really good at doing the most mundane work like tweaking syntax, analyzing structure and optimizing functional snippets of code. They are doing great at increasing the productivity of architect-level coders who have much to think about and too little time to actually grind out complex code.
My last take is controversial and questionable. I feel that there will always be a human element required to put together requirements and formulate overall architectures that will produce useful functionality that solves problems. Chatbots are not good at this because it requires being visionary to a certain extent and that is a far more complex nuance to problem solving. I don't think a chatbot is near that level as it is essentially AGI.
Lastly, AudreyH's mention of testing is relevant. I see it in my current workplace and people are using chatbots to write tests, to enhance existing tests and to make pipelines are more robust and reliable, reducing the iterations required to reduce defects. I think the fact that testing is so unglamourous and so mundane that it gets largely ignored
EDIT:
Oh yeah, my takes here are arrogant and condescending but this is an anonymous forum so I can vent about this freely here. I don't air these views publicly at the workplace because it comes off as "grumpy old man" stuff. I just keep my S/W religion and politics to myself at work and professionally and I think the younger engineers respect me for this.
surprising
Recycles dryer sheets
- Joined
- Feb 7, 2023
- Messages
- 365
Talking to a colleague recently who is very against AI in the creative space because it is stealing the work of artists, but is OK with AI as a coding assistant (he is a software developer).
I think the analogy would be something like - AI as a coding *assistant* is OK, just as AI as an artist *assistant* is beneficial. However, AI replacing the artist is not OK, and AI replacing the software architect/engineer is not OK. AI is not there yet for completely replacing software architecture and development, but eventually it will be. For some things, it is already there. Kids are using AI to completely develop Roblox games. The kids have 0 software experience, they just take the resulting code and cut-n-paste it into Roblox and iterate. I actually think people with 0 understanding of computers and software will have an advantage because they won't have an inherent bias when interacting with the AI.
I think the analogy would be something like - AI as a coding *assistant* is OK, just as AI as an artist *assistant* is beneficial. However, AI replacing the artist is not OK, and AI replacing the software architect/engineer is not OK. AI is not there yet for completely replacing software architecture and development, but eventually it will be. For some things, it is already there. Kids are using AI to completely develop Roblox games. The kids have 0 software experience, they just take the resulting code and cut-n-paste it into Roblox and iterate. I actually think people with 0 understanding of computers and software will have an advantage because they won't have an inherent bias when interacting with the AI.
Route246
Full time employment: Posting here.
- Joined
- Jun 22, 2023
- Messages
- 874
The cat is out of the bag and we have to take the good with the bad. There is no turning back and people need to accept that there will be some negative consequences. Acceptance is a tough pill to swallow if it directly impacts you. During my 50+ year career as a S/W engineer I was obsoleted out of jobs on a regular basis due to technology changes. I also had enough fear to do something about this and I was always developing skills with new technologies, new languages, new environments and I was able to do pretty well, much better than some ex-colleagues who refused to exit their comfort zones and ended up getting aged out into unwanted and early retirement or forced career changes. Considering where I started writing 360/370 Assembler code to using coding chatbots today to write complex object oriented algorithms is something I'll look back on with great pride. In a field where it is extremely easy to become a dinosaur I managed to avert that consequence and will retire in October at peak earnings, leaving unvested 7-figures on the table and with a smile on my face without regrets. But, what I am most proud of, and I was discussing this privately with an older peer like myself just last week, is I never reverted to the "grumpy old man" character and always kept those opinions to myself and spent more time listening to young, energetic, highly intelligent young S/W engineers and provided support and reinforcement. This is for selfish reasons, too, as I am a shareholder and I want these young engineers to increase my shareholder value when I'm gone.
The thing that gets lost in this debate is that person-generated code is pretty garbage. AI-generated code is always compared to what “a skilled programmer can accomplish”. There are not many of those. The tech market has grown so much that it is flooded with people who are just not very good. The truth is AI is a better than average programmer. People who complain about AI occasionally generating nonsensical code or making changes in completely unrelated areas of code to the task at hand have never been responsible for reviewing code from junior and mid level programmers.
It’s not surprising to me that AI generates bad code sometimes. The AI was trained on all the garbage code written by humans!
If you’re a country like India that relies on outsourced repetitive/mundane code activities, or if you’re a programmer early in your career, I would be thinking about diversification.
If you’re a senior programmer, you’re about to be way more productive and make way more money.
- Joined
- Oct 13, 2010
- Messages
- 12,058
I really meant to post this [ AI and the future ] in this thread, but it's been quoted over there, so I'll leave it, but just mention the post here.
There's a an hour-long read about how AI might play out in the next couple of years https://ai-2027.com/
That includes some predictions based on how fast AI coding has progressed so far, doubling in size of project every few months. The guy, Kokotajlo, has AI beating a human coder by 2027.
"....but people who know how to manage and quality-control teams of AIs are making a killing"
There's a an hour-long read about how AI might play out in the next couple of years https://ai-2027.com/
That includes some predictions based on how fast AI coding has progressed so far, doubling in size of project every few months. The guy, Kokotajlo, has AI beating a human coder by 2027.
The AI 2027 story says:
"....but people who know how to manage and quality-control teams of AIs are making a killing"
Last edited:
Hmm... I tell folks the exact opposite! I had 6 xmas layoffs in STEM and tell everyone to avoid it as the costs and training and treadmill are not worth the pay and constant job hunting with career interruptions and restarts. You never pay for the college or grad school opportunity cost. I originally took up LinkedIn to keep track of the places I had worked. Could not remember all the layoffs.
The skills are absolutely necessary. I consider STEM to be a 1928 liberal arts degree. Pays nothing but a good foundation. My issue is that the expense and delay of getting college credentials impairs both learning and earning.
Ah what a difference a few years makes. I was a freshman in college when my dad supplemented my high school heathkit tv as a monitor with the 99/4a and a tape deck for data storage. I did a Runge-Kutta program in basic. You were fascinated, I was bored stiff. Only when I was finishing grad school and did a few reports in the Mac lab with printers and software that worked did I get a glimpse of the future.
I just finished reading Desktop Engineering pdf on AI and simulation in engineering. Color me unimpressed.
My son is a software architect (not sure of what that is exactly) in a huge national company. I talked to him about it. He will be 51 soon and hopes to last at work until 55, then retire. He says for the type of complex applications they design for, they still need skilled software design/developers. He sees the value in that AI might be able to do the boring coding that people hate to do, then so then make mistakes that he ends up spending days trying to find and correct. But it’s new world and he’s not sure what will happen. There are so many variables.
Sunset
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
Over Age 50 is about when companies start to think a Software developer is too old or doesn't know the new stuff, and they can hire new grads for cheaper. It's pretty brutal.
WWDog
Recycles dryer sheets
You hire yourself out as the guy who cracks the whip on the assembly line. Crude analogy but relevant to the direction of where things are moving.
Last edited:
- Joined
- Oct 13, 2010
- Messages
- 12,058
There will be the new job of making sure the AI is doing what it claims to be doing.
At first that will just take human skills, but as things progress, it will be beyond human capability. At that point, the job will be to manage AI's by people who know how to operate more narrow AI's that are tasked with monitoring the more general AI's.
At first that will just take human skills, but as things progress, it will be beyond human capability. At that point, the job will be to manage AI's by people who know how to operate more narrow AI's that are tasked with monitoring the more general AI's.
Route246
Full time employment: Posting here.
- Joined
- Jun 22, 2023
- Messages
- 874
Absolutely agree. Managed to dodge that bullet (retiring at 68 in October) by being self-aware, fearful of being aged out and continuously venturing out of my high paying comfort zone. It worked, as many ex-colleagues who did get aged out thought I was crazy trying to learn new things and felt like I was putting myself through unnecessary pain and torture. I got the last laugh and walking away with a smile on my face.
One really important thing was avoiding the "old fart" language that is so easy to fall into. Nobody wants to hear an old fart reminisce about "back in the day" we did this or that. I never talk about the past and have laser focus on the future. The current hot topic is AI-assisted coding which is really one of those once in a generation quantum leaps that will leave many young and old dinosaurs behind.
target2019
Give me a museum and I'll fill it. (Picasso) Give me a forum ...
Whipper Snapper?
Wow, I'm sorry to hear that. I was lucky in that they taught the 'C' language in college and I was able to find a good job right out of school. Moved to Java in the late 90's and was able to code in that until the 2015. I could have stayed with that, but dabbled in the functional languages (Scala).
I never worried about a layoff. But I might have gotten out at the right time.
I'll share my (modest) experience with grok.
Let me give you some context
1/ I'm not a developer. I w0rked for high tech hardware and software companies making complex systems / (B2B) solutions. My j0b was to specify workflow, use cases, features, test cases, user experiences, interfaces... etc... for end to end solutions integrating our products and 3rd party products in a global solution. Someone else was developing what was missing in our products...
2/ I kinda learnt BASIC in the mid-80s (self-taught), PASCAL + C + Assembly in the mid-90s (school)
a/ I was never good at it
b/ I never liked it
c/ Until recently, I might have produced less than 100 lines of code in the span of 40 years...
3/ For one of my hobby; I'm using a software tool which has a built in LUA SDK.
a/ 2-3 years ago, I kind of self-taught myself LUA (mostly searching the internet and Stackoverflow)
b/ 3 days ago I decided to try grok for the first time to help me with parsing a bitmap (several variables with different bit start and bit length, on the same message). It took me 3 prompts to clarify what I needed (not parsing a bitmap picture LOL!), then grok provided the code. I used it and it worked. Great, it saved me time learning (or searching Stackoverflow) the technically of bit32.rshift and bit32.band.
c/ However, I wanted to used the code in a loop. Grok provided 2 (technical) cases in the implementation. That seems 'silly' to me. I don't know how other developers think about it (and again I'm not a developer), but I want my code to be as compact as possible with as few (technical) cases as possible; so it's efficient and easily readable (that's one of my goal when I write LUA for this SDK). So I re-thought about it and managed to use 1 technical case (1 line of code) in my loop. Voila.
d/ Maybe I should have specified to grok that I want to use the code in a loop, and it would have provided me with a more efficient solution, IDK...
Conclusion
I/ For me (as a non developer), it's a fantastic tool: I provide / specify a functional problem, grok solves it technically
II/ Is the code provided by grok the most efficient (compact - in my case) possible?
Let me give you some context
1/ I'm not a developer. I w0rked for high tech hardware and software companies making complex systems / (B2B) solutions. My j0b was to specify workflow, use cases, features, test cases, user experiences, interfaces... etc... for end to end solutions integrating our products and 3rd party products in a global solution. Someone else was developing what was missing in our products...
2/ I kinda learnt BASIC in the mid-80s (self-taught), PASCAL + C + Assembly in the mid-90s (school)
a/ I was never good at it
b/ I never liked it
c/ Until recently, I might have produced less than 100 lines of code in the span of 40 years...
3/ For one of my hobby; I'm using a software tool which has a built in LUA SDK.
a/ 2-3 years ago, I kind of self-taught myself LUA (mostly searching the internet and Stackoverflow)
b/ 3 days ago I decided to try grok for the first time to help me with parsing a bitmap (several variables with different bit start and bit length, on the same message). It took me 3 prompts to clarify what I needed (not parsing a bitmap picture LOL!), then grok provided the code. I used it and it worked. Great, it saved me time learning (or searching Stackoverflow) the technically of bit32.rshift and bit32.band.
c/ However, I wanted to used the code in a loop. Grok provided 2 (technical) cases in the implementation. That seems 'silly' to me. I don't know how other developers think about it (and again I'm not a developer), but I want my code to be as compact as possible with as few (technical) cases as possible; so it's efficient and easily readable (that's one of my goal when I write LUA for this SDK). So I re-thought about it and managed to use 1 technical case (1 line of code) in my loop. Voila.
d/ Maybe I should have specified to grok that I want to use the code in a loop, and it would have provided me with a more efficient solution, IDK...
Conclusion
I/ For me (as a non developer), it's a fantastic tool: I provide / specify a functional problem, grok solves it technically
II/ Is the code provided by grok the most efficient (compact - in my case) possible?
